Background: Studies have suggested that there is a significant association between social engagement and depression symptoms. However, this association may differ in people with different features such as different sociodemographic characteristics and health conditions.
Methods: Research data were obtained from the CHARLS database. The causal inference was performed with the propensity score. We used the linear mixed-effects model tree algorithm under the causal inference frame for subgroup identification analysis.
Results: We included 13,521 participants, and the median follow-up time is 4 years. Under the casual inference frame, the association between social engagement and depression symptoms is confirmed for all included individuals (OR = 0.957, P = 0.016; 95%CI: 0.923-0.992). Using the linear mixed-effects model tree, we found two subgroups, including middle-aged and elderly residents who live in rural areas with <6 h of sleep and those living in urban areas, could benefit more from social engagement. After using the propensity score method, all the two subgroups selected are statistically significant (P = 0.007; P = 0.013) and have a larger effect size (OR = 0.897, 95%CI: 0.830-0.971; OR = 0.916, 95%CI: 0.854-0.981) than the whole participants. As for sex difference, this associations are statistically significant in male (OR: 0.935, P = 0.011, 95%CI: 0.888-0.985) but not in female (OR: 0.979, P = 0.399, 95%CI: 0.931-1.029).
Conclusions: Our findings indicate that social engagement may reduce the risks of depressive symptoms among all individuals. The identified subgroups of middle-aged and elderly residents who live in rural areas with <6 h of sleep and those who live in urban areas may benefit more from the social engagement than the whole participants.
Keywords: causal inference; depressive symptoms; middle-aged and elderly Chinese; social engagement; subgroup identification.
Copyright © 2022 Yang, Li, Zhao, Wang, Mi, Pei, Zhao and Chen.